Bilingual Neural Machine Translation From English To Yoruba Using A Transformer Model

Abstract

The necessity for language translation in Nigeria arises from its linguistic diversity, facilitating effective communication and understanding across communities. Yoruba, considered a language with limited resources, has potential for greater online presence. This research proposes a neural machine translation model using a transformer architecture to convert English text into Yoruba text. While previous studies have addressed this area, challenges such as vanishing gradients, translation accuracy, and computational efficiency for longer sequences persist. This research proposes to address these limitations by employing a transformer- based model, which has demonstrated efficacy in overcoming issues associated with Recurrent Neural Networks (RNNs). Unlike RNNs, transformers utilize attention mechanisms to establish comprehensive connections between input and output, improving translation quality and computational efficiency.